Organization of the Olfactory Pathway and Odor Processing in the Antennal Lobe of the Ant Camponotus ﬂoridanus

CHRISTINA ZUBE,1 CHRISTOPH JOHANNES KLEINEIDAM,1 SEBASTIAN KIRSCHNER,2 JAKOB NEEF,1 AND WOLFGANG RO¨ SSLER1* 1Department of Behavioral Physiology and Sociobiology, Biozentrum, University of Wu¨rzburg, Wu¨rzburg, Germany 2Department of Developmental and Comparative Psychology, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

ABSTRACT Ants rely heavily on olfaction for communication and orientation. Here we provide the first detailed structure–function analyses within an ant’s central olfactory system asking whether in the carpenter ant, Camponotus floridanus, the olfactory pathway exhibits adap- tations to processing many pheromonal and general odors. Using fluorescent tracing, confocal microscopy, and 3D-analyses we demonstrate that the antennal lobe (AL) contains up to Ϸ460 olfactory glomeruli organized in seven distinct clusters innervated via seven antennal sensory tracts. The AL is divided into two hemispheres regarding innervation of glomeruli by either projection neurons (PNs) with axons leaving via the medial (m) or lateral (l) antenno- cerebral tract (ACT). M- and l-ACT PNs differ in their target areas in the mushroom-body calyx and lateral horn. Three additional ACTs project to the lateral protocerebrum only. We analyzed odor processing in AL glomeruli by retrograde loading of PNs with Fura-2 dextran and fluorimetric calcium imaging. Odor responses were reproducible and comparable across individuals. Calcium responses to pheromonal and nonpheromonal odors were very sensitive (10Ϫ11 dilution) and patterns were partly overlapping, indicating that processing of both odor classes is not spatially segregated within the AL. Response patterns to the main trail- pheromone component nerolic acid remained stable over a wide range of intensities (7–8 log units), while response durations increased indicating that odor quality is maintained by a stable pattern and intensity is mainly encoded in response durations. The structure–function analyses contribute new insights into important aspects of odor processing in a highly advanced insect olfactory system.

Indexing terms: glomeruli; 3D-reconstruction; calcium imaging; projection neurons; mushroom bodies; lateral horn; insect brain; antenno-cerebral tract

Odors play an essential role for the regulation of social interactions and colony organization in social insects like ants or bees. Ants heavily depend on chemical communication and species-specific pheromones are essential to This article includes Supplementary Material available via the Internet organize social behavior (Ho¨lldobler and Wilson, 1990). at http://www.interscience.wiley.com/jpages/0021-9967/suppmat. For example, trail-following behavior, alertness, recruit- Grant sponsor: German Science Foundation DFG; Grant number: SFB ment, or signaling of the reproductive state are coordi- 554 (A6 and A8); Grant sponsor: Evangelisches Studienwerk e.V. Villigst. nated by the action of pheromones (Ho¨lldobler, 1995). In *Correspondence to: Wolfgang Ro¨ssler, Department of Behavioral Phys- addition, substances on the body surface (cuticular hydro- iology and Sociobiology, Zoology II, Biozentrum, University of Wu¨rzburg, Am Hubland, 97074 Wu¨rzburg, Germany. carbons) serve as intra- and interspecific recognition cues E-mail: [email protected] affecting nestmate recognition as well as intra- and inter- species aggressive interactions (e.g., Singer, 1998; Lenoir et al., 1999; Lahav et al., 1999). Besides signals used for 426 species-specific communication, a large variety of environ- male AL (e.g., Hansson et al., 1991; Vickers et al., 1998; mental odors play an important role for orientation and Hansson and Anton, 2000; Rospars and Hildebrand, the location and evaluation of food sources (Ho¨lldobler and 2000). Sex-pheromone-specific macroglomeruli were also Wilson, 1990). Thus, the detection, processing, and recog- described in the male cockroach and in the honeybee nition of a remarkable number of chemical cues by the drone (Malun et al., 1993; Brockmann and Bru¨ckner, olfactory system are essential for the survival and repro- 2001; Sandoz, 2006). A study in leaf-cutting ants revealed ductive success of ant colonies. These chemosensory tasks a macroglomerulus in large workers of Atta sexdens and A. require sophisticated sensory machinery and advanced vollenweideri obviously not associated with sex- olfactory neuronal networks in the brain of each individ- pheromone processing (Kleineidam et al., 2005). Interest- ual. ingly, the macroglomerulus was absent in small workers, In insects the antennal lobes (ALs) in the brain are the indicating caste-specific differences in AL organization in first relay station for processing of olfactory information leaf-cutting ants. received by olfactory receptor neurons (ORNs) housed in Only a few studies have investigated processing of non- sensilla on the antennae, the main olfactory receptor or- sexual pheromones in insects. In social insects nonsexual gans. ORN axons terminate in olfactory glomeruli within pheromones are essential for colony organization and sur- the AL. Glomeruli are spheroidal regions of dense synap- vival. Calcium-imaging studies in the carpenter ant Cam- tic neuropil and in most systems studied so far glomeruli ponotus rufipes and in the honeybee did not reveal a spe- can be regarded as functional units in odor processing cific clustering of specialized glomeruli responsive to, e.g., (e.g., Hildebrand and Shepherd, 1997; Hansson and An- alarm pheromone, and the same glomeruli or directly ton, 2000; Galizia and Menzel, 2001; Sandoz, 2006). Mo- neighboring glomeruli were shown to participate in re- lecular and tracing studies have shown that in both ver- sponses to nonpheromonal odors (Galizia et al., 1999b,c; tebrates and insects axons from ORNs that express the Sachse et al., 1999; Sandoz, 2006). A recent electrophysi- same odorant receptor or that have a similar response ological study in ants showed that alarm-pheromone re- profile converge on the same glomeruli (Vassar et al., sponsive PNs innervate a specific cluster of normally sized 1994; Mombaerts et al., 1996; Ro¨ssler et al., 1999a,b; glomeruli within the AL, indicating at least some degree of Vosshall et al., 2000; Xu et al., 2000; Carlsson et al., 2002; anatomical segregation of pheromone processing in the Wang et al., 2003). After local processing by a network of ant AL (Yamagata et al., 2006). local interneurons, olfactory information is transferred via The focus of the present study was to ask whether the projection neurons (PNs) to higher integration centers in olfactory system of the highly olfactory carpenter ant, the protocerebrum, the mushroom bodies (MBs), and ol- Camponotus floridanus, expresses specific neuroanatomi- factory neuropils in the lateral protocerebral lobe (LPL), cal and/or neurophysiological specializations according to most prominently within the lateral horn (LH). In the its elaborated role in dealing with a large number of honeybee the output from glomeruli is relayed in a distinct pheromonal and nonpheromonal odors. In particular, we pattern to the MB and LH via two prominent antennoc- wanted to test: erebral tracts (medial [m] and lateral [l] ACT) formed 1) whether the general organization, the number, and the mainly by uniglomerular PNs and three small mediolat- input–output connections of olfactory glomeruli in the eral ACTs to the LPL mainly formed by multiglomerular AL differ from those found in the well-investigated PNs (Abel et al., 2001; Mu¨ller et al., 2002; Kirschner et al., honeybee (Galizia et al., 1999b; Kirschner et al., 2006); 2006). Most important, the dendritic fields of the m- and 2) whether the AL has segregated regions for processing l-ACT PNs are separated, forming two hemispheres of pheromonal and nonpheromonal odors; and glomeruli in the AL. Their target areas in the MBs and LH 3) whether the ant olfactory system is specialized to de- also remain largely segregated, indicating that olfactory tect minimal quantities of pheromonal signals and, at information from glomeruli in two AL hemispheres is the same time, is able to respond appropriately to large transferred and processed via two separate uniglomerular ranges of intensities as, for example, required for trail PN output channels (Kirschner et al., 2006). pheromone detection. The functional aspects of odor processing in the hymenopteran brain has been mainly approached by in vivo To answer the first question we used fluorescent tracing optical imaging studies in the honeybee showing that techniques, confocal microscopy, and 3D reconstructions within the AL different odors elicit odor specific glomeru- to analyze the anatomical organization of AL glomeruli, lar activation patterns depending on odor properties such their antennal sensory input, and their protocerebral out- as molecule identity, intensity, and the composition of put connections. To answer the second and third questions mixtures (Joerges et al., 1997; Galizia et al., 1999b; Sachse we retrogradely loaded PNs with a calcium-sensitive dye et al., 1999; Carlsson et al., 2002; Carlsson and Hansson, (Fura-2 dextran) and performed fluorimetric calcium- 2003; Sachse and Galizia, 2003; Sandoz, 2003, 2006; Peele imaging of odor responses of the glomerular (PN) output et al., 2006). Despite the great importance of olfaction in in the AL. For the third question we focused on intensity ants, up to now only one imaging study was carried out coding in response to nerolic acid, a major trail pheromone (Galizia et al., 1999a), and structural data on the detailed component in C. floridanus. organization of the olfactory pathway is rather limited (Gronenberg, 1999, 2001; Hoyer et al., 2005; Kleineidam et al., 2005). MATERIALS AND METHODS Research on pheromone processing in insects has largely focused on sex pheromones, especially in moths Animals (e.g., Christensen and Hildebrand 1987; Hildebrand and Workers of the carpenter ant, C. floridanus, were used Shepherd, 1997), where sex-pheromone-specific glomeruli for the experiments. Ants were taken from a colony with a form a macroglomerular complex at the entrance of the singly mated founding queen collected in Orchid Island, 427

FL. Only large workers (head width Ͼ3 mm, body length (P2000, Sutter Instruments, Novato, CA) using borosili- Ϸ10 mm) were used for both neuroanatomical and neuro- cate capillaries (1B100F-3, Precision Instruments, Sara- physiological experiments. The colony was reared under sota, FL). The broken tip of each glass electrode was constant conditions in an environmental chamber at 25°C coated with small dye crystals (Ϸ200 ␮m in diameter). and 85% humidity on a 12/12-h photoperiod. Animals were Prior to inserting the dye-coated pipettes into brain tissue fed twice a week with honey water, bathkar, and dead the region of interest was carefully perforated with an cockroaches. Fresh water was given once a week. Ants unbroken glass pipette. Subsequently, the dye electrode were anesthetized with CO2 and prepared for anatomical was plunged into the perforated area and remained for up and physiological experiments as described below. to 10 seconds in the target area. The pipette was removed Neuroanatomical procedures and the brain immediately rinsed with fresh ant-ringer solution to wash out excessive dye. To investigate the AL Whole-mount preparations. For visualization of connections within the protocerebrum, dye was inserted brain structures, especially AL glomeruli, we utilized tis- during the anterior preparation directly into the AL neu- sue autofluorescence caused by glutaraldehyde fixation. ropil. For double labeling of the projection areas of the The ants were decapitated and the head capsule fixed in medial and lateral ACT in the MBs and LH we used a dental wax-coated dishes. The head capsule was opened technique described by Kirschner et al. (2006), but access- by cutting a window between the compound eyes and the ing the ant brain from posterior. The m-ACT was labeled brain was rinsed immediately using fresh ant-saline solu- by insertion of the dextran tracer in the medial-caudal tion (127 mM NaCl, 7 mM KCl, 1.5 mM CaCl2, 0.8 mM protocerebrum and for the l-ACT the dextran tracer was Na2HPO4, 0.4 mM KH2PO4, 4.8 mM TES, 3.2 mM Treha- placed into the lateral-caudal protocerebrum. To double- lose, pH 7.0). Glands and tracheae were removed and the label ORN axons and the dendritic arborizations of PNs in brains were dissected out and fixed immediately in cold the AL we combined anterograde ORN mass fills with 1% glutaraldehyde in phosphate-buffered saline (PBS, pH retrograde mass fills of m- and l-ACT PNs in a posterior 7.2) for 4 days at 4°C. The brains were then washed in preparation. Briefly, one antenna was cut at the base of PBS (5 ϫ 10 minutes) and dehydrated in an ascending the scapus and dextran-conjugated fluorescent dye was series of ethanol (30%, 50%, 70%, 90%, 95%, 3 ϫ 100%, 10 immediately applied at the cut surface. In a second step minutes each step). Finally, the brains were cleared in either the m- or l-ACT was counterstained using the methylsalicylate (M-2047, Sigma Aldrich, Steinheim, Ger- method described above. many) and mounted in special aluminum slides with a central hole covered by thin coverslips from both sides. Confocal laser-scanning microscopy and 3D- Whole-mounts were stored at Ϫ20°C. reconstructions of glomeruli and tracts Tracer application and tissue preparation. For antennal backfill preparations, all animals were fixed in All brains were viewed as whole-mount preparations customized acrylic blocks and the heads and antennae using two confocal laser scanning microscopes (Leica TCS were stabilized with dental wax. Both antennae were cut LP and Leica TCS SP2 AOBS; Leica Microsystems, Wet- at the lower part of the pedicellus. The cut surface was zlar, Germany) equipped with an argon/krypton and immediately covered with a drop of dextran-biotin (D- helium/neon laser. Excitation wavelengths were 568 nm 7135, Molecular Probes, Eugene, OR) dissolved in distilled for rhodamine and streptavidin-conjugated Alexa 568 and water, and the preparation was kept in a moistened cham- 488 nm for Alexa Fluor 488. Two different HC PL APO ber overnight to let the dye diffuse. The brains were then objective lenses were used for image acquisition (10 ϫ 0.4 dissected and fixed immediately in 4% formaldehyde in NA imm and 20 ϫ 0.7 NA imm) and optical sections were 0.1 M PBS overnight at 4°C. After washing with 0.1 M taken at distances between 1–10 ␮m. In certain cases a PBS (3 ϫ 10 minutes) the brains were incubated in Alexa digital zoom of 2–3ϫ was applied. Double-labeled speci- 568-conjugated streptavidin (S-11223, Molecular Probes) mens were scanned sequentially. All confocal image in PBS with 0.2% Triton-X (1:250), first for 1 hour at room stacks were viewed and processed with the 3D- temperature, then overnight at 4°C. Finally, the brains reconstruction software AMIRA 3.1 (Mercury Computer were washed with 0.1 M PBS (4 ϫ 5 minutes on a shaker Systems, Berlin, Germany). To obtain a better signal-to- at room temperature), dehydrated in an ascending alcohol noise ratio some preparations were deconvoluted using series (30%, 50%, 70%, 90%, 95%, 3 ϫ 100%, 10 minutes the AMIRA deconvolution algorithm. In doubly labeled each step) and cleared in methylsalicylate and mounted in preparations each channel was set to a false color. Images permount (Fischer Scientific, Schwerte, Germany). were scaled and snapshots were taken from single optical For labeling of the ACTs the ants were fixed in dental sections or complete stacks. Screenshots were further pro- wax dishes. The heads were fixed in an anterior and cessed in Adobe Photoshop 6.0 and 7.0 software (Adobe posterior preparation to access the front or back of the Systems San Jose, CA) and adjusted for brightness and brain depending on the area of dye application. In either contrast. Anatomical directions refer to Strausfeld (2002) case a cut between the compound eyes was made to open and Kirschner et al. (2006). Single glomeruli were clearly the head capsule and glands and tracheae were removed. visible as densely packed neuropil structures and were The tracer was applied into different brain regions to outlined in all focal planes (yz,xy,xz). 3D-reconstructions selectively stain the tracts of interest (Kirschner et al., of individual glomeruli were done using the Amira 3.1 2006). Two dextran conjugates were used: rhodamine dex- feature “wrap.” Antennal sensory input tracts as well as tran with biotin (3,000 MW, lysine-fixable; Microruby, D AL output tracts (and somata) were rendered using the 7162, Molecular Probes) and Alexa Fluor 488 dextran “interpolate” and “automatic threshold” feature. Individ- (10,000 MW, lysine-fixable; D 22910, Molecular Probes). ual glomeruli or cluster of glomeruli associated with the The tracers were applied as follows: glass micropipettes same antennal sensory input tracts (T1–T7) were color were pulled with a horizontal laser-electrode puller coded. 428

Preparation and dye loading for calcium quently applying an N ϫ N filter (5 ϫ 5 pixels with 5 imaging iterations) to reduce noise. Autofluorescence and stained neurons caused inhomogeneous fluorescence images Ants were briefly (for a few seconds) anesthetized with (background fluorescence) and by subtracting the average CO2 and fixed in a Plexiglas stage using soft dental wax ratio-image using frames 1 to 9 from all ratio images the (surgident periphery wax, Heraeus Kulzer, Germany). background was set to zero prior to odor stimulation. Compared to treatment with CO2, anesthetization by cool- Filtered ratio-images with background subtraction are la- ing on ice did not show any obvious differences in the ⌬ beled as (F340/F380). responses to the odors tested (see below). A small window Following odor stimulation, calcium signals were mea- was cut into the head capsule to access the ventral part of ⌬ sured as changes in fluorescence (F340/F380) and consid- the brain and the site for dye application. A sharp glass ered as neuronal activity in response to a given odor electrode coated with a few crystals of Fura-2 dextran stimulus when they exceeded 40% of the maximal re- (potassium salt, 10,000 MW, F3029, Molecular Probes) sponse. The maximal response was measured across all dissolved in 2% bovine serum albumin solution was in- different odors and concentrations tested in the prepara- serted for several seconds into the lateral protocerebrum, tion. Neuronal activities (activity areas) are represented dorsolaterally to the vertical lobe of the MBs, aiming for as false-color-coded images using the average of frame 11 the m- and l-ACT (see method above and Fig. 2F,G). Fol- to 14 (during odor stimulation). In most cases, activity lowing dye application the brain was rinsed with ant- areas could not be assigned to single glomeruli because saline solution to remove excess dye. The window in the often their size was considerably larger than the size of head capsule was closed with the cut piece of cuticle and single glomeruli. The circular shape and small size (10–40 the ants were released from the Plexiglas stage for 4 ␮m) of other activity areas were similar to the dimensions hours. During the staining period the ants were allowed to of single glomeruli and termed activity spots. move freely in a moistened Plexiglas container before they The spatial overlap of activity areas in response to two were fixed again in the Plexiglas stage as before. Anten- odors was calculated as the percentage of an odor specific nae were immobilized with wax and a larger window was activity area that was activated by the other odor. All cut into the head capsule to access the whole brain and the pixels within the AL with intensity values above threshold ALs. Glands and tracheae were carefully removed and the (40% of the maximal response) during stimulation with esophagus was cut at the mouth parts and pulled out of one odor were counted and compared with the odor- the head capsule to prevent movement of the brain during specific activity areas elicited by one of four other odors data acquisition. We prepared a total of 307 ants for using the software AMIRA 3.1 (Mercury Computer Sys- Fura-2 dextran labeling and calcium imaging; 145 (47%) tems, Germany). Calcium signals in a total of five animals Ϸ resulted in bright staining in the ALs, but 82 ( 27%) of and in response to five different odors were used for this the ants with staining in the ALs showed no spontaneous analysis. activity or responses to odor stimulation. In 34 ants For four different odors, threshold odor concentrations Ϸ ( 11%) clear spontaneous activity was observed, but no were measured in 4–11 animals using the threshold cri- responses to odor stimulation. We therefore used 29 ants terion mentioned above. In order to analyze the dynamic Ϸ ( 9–10% of the total number) that were clearly stained, range of the calcium responses at the different odor con- showed spontaneous activity, and clear responses to odor centrations, at least two areas with the highest calcium- stimulation for further analyses. We were able to test the signal amplitudes (circular regions of interest, ROIs) were complete stimulation program including all odor intensi- selected and normalized within each animal. The calcium- Ϸ ties in seven ( 2%) of these 29 ants (see below). signal amplitude within the ROIs and the duration (number of frames) was measured across all tested odor Imaging concentrations (dilutions: 10Ϫ11,10Ϫ8,10Ϫ5, and 10Ϫ2). Calcium-imaging experiments were performed using an To analyze whether signal amplitude and duration cor- Olympus imaging system (Cell, v. 1.1-2.3) with an upright relate with odor intensity, two ROIs of each preparation epifluorescent microscope (BX51WI; with the filter set were separated according to either showing clear UM2FUR) equipped with an LD 20ϫ water-immersion concentration-independent calcium-signal amplitudes lens (XLUMP, NA 0.95) and epifluorescent illumination (cdϪ) or concentration-dependent calcium-signal ampli- was provided by a 150 W xenon light source (MT20, with tudes (cdϩ). Signal durations within the resulting two excitation filters for 340 nm and 380 nm). The focal plane classes of ROIs were compared using a Spearman’s- within the AL was adjusted to a depth of 35 ␮m below the rank correlation on pooled data of seven animals. surface of the AL using a piezo-driven nanofocusing system (PIFOC, PI, Germany). For each stimulus a series of Odor stimulation 24 double frames was recorded with an air-cooled CCD A constant and moistened air stream of 1 L/min was camera (model 8484-03G, Hamamatsu Photonics, Japan) produced by two independent flow controllers (VC-2LPM, at a sampling rate of 4 Hz. A 2 ϫ 2 on chip binning Alicat Scientific, Tucson, AZ), both set to 0.5 L/min and resulted in an image pixel size of 0.645 ϫ 0.645 ␮m. connected to two solenoid valves controlled by the imaging Exposure times ranged from 40–60 ms for the first frame software. The solenoid valves allowed switching of each of at 340 nm and 20–30 ms for the second frame at 380 nm. the two flow channels through a plastic cartridge (1 mL) Odor stimulation started at frame 10 and was terminated containing a filter paper (1 cm2). Only one of the two flow at frame 14 (1 second). Repeated odor stimuli were given channels was used in the experiments. For odor stimula- at an interstimulus time interval of at least 1 minute. tion, 5 ␮L of the odor diluted in mineral oil (Sigma Aldrich, Imaging data were analyzed by calculating the ratio of Deisenhofen, Germany) was applied onto the filter paper fluorescence intensity in the images taken at 340 and 380 and the cartridge was placed into the olfactometer. Odor ϭ Ϫ1 Ϫ12 nm excitation for each pair as: R F340/F380, and subse- dilutions ranged from 10 to 10 and experiments al- 429 ways started with the lowest odor concentration. As con- side of the AL. The relatively large T4 proceeded trans- trol stimulus, 5 ␮L of mineral oil was applied onto the versally through the inner lateral part of the AL (Figs. filter paper. 1C,D, 2C). Opposite to this cluster, in the dorsal-medial Eight different odors including two pheromones were AL, Ϸ36 glomeruli were innervated by T5, which pro- used for stimulation. As general odors, citral, isoamylac- ceeded transversally through the AL in a dorsal-medial etate (IAA), nerol, 1-hexanol, heptanal, and 1-octanol (all direction. Next to the T5 cluster we found the largest from Sigma Aldrich) were used. As pheromones of C. flori- group of glomeruli in the AL, the T6 cluster, consisting of danus, the alarm pheromone, n-undecane (Sigma Aldrich) Ϸ128 comparatively small glomeruli (Table 1; Figs. 1B–D, and the trail pheromone (releaser component), nerolic acid 2C). The T6 proceeded along the dorsal-medial surface of (Haak et al., 1996; Cardiff Chemicals, Cardiff, UK) were the AL and entered its target glomeruli from the periph- used for odor stimulation. ery. The smallest cluster of T7 glomeruli was located in Activity areas were described for all odors, but not all the dorsal part of the AL, close to the DL, and consisted of odors were tested in each experiment. Odor representa- six comparatively large glomeruli. The volume of these tion across animals was investigated using heptanal, glomeruli was calculated as 4–10 times larger than that of isoamylacetate, citral, and nerolic acid. The overlap of most other AL glomeruli, and their innervation was char- activity areas in response to different odors and the acterized by particularly brightly labeled bundles of axons threshold concentration for each single odor were investi- (Figs. 1D, 2C, thin dashed lines; Table 1; and Suppl. 3D gated using nerolic acid, n-undecane, citral, heptanal, and online material). In all preparations these glomeruli were 1-octanol. The duration of the calcium signal across dif- stained brightly. The corresponding ORN axons reached ferent odor concentrations was investigated for the known the T7 glomeruli via several loose bundles which we term trail pheromone component nerolic acid. the T7 tract. Hemispherical division of AL glomeruli by RESULTS m- and l-ACT output Two prominent AL output tracts, the m- and l-ACT, Sensory-tract innervation of glomeruli were successfully double-labeled in two preparations and The mass-labeled projections of ORN axons were ana- the m-ACT was selectively labeled in four cases, the l-ACT lyzed to investigate the pattern of sensory-tract-specific in three cases. One example of a doubly labeled AL and innervation of glomerular clusters in the AL of C. florida- the corresponding 3D-reconstruction of glomeruli supplied nus (n ϭ 6). A representative example is shown in Figure by the m-ACT (magenta) and l-ACT (green) are shown in 1 and the corresponding 3D-reconstruction is shown in Figure 2D,E. Please note that the colors in 2D,E do not Figure 2B,C (see also Suppl. 3D online material). The reflect connectivity with glomeruli with similar colors in complete brain reconstruction in Figure 2A shows that C. Fig. 2B (for a summary of the input-output connectivity, floridanus has small optic ganglia and relatively large ALs see Table 1). Our recent study in the honeybee revealed a underlining the mainly olfactory sensory orientation of similar hemispherical division of AL glomeruli (Kirschner this ant. In the AL of six preparations we counted between et al., 2006). Compared to the situation in the honeybee, 434 and 464 glomeruli, indicating some variability in the however, the two AL hemispheres in C. floridanus con- total number of glomeruli in large workers. The mean tained a much higher number of glomeruli (Table 1) and number of glomeruli was 452 Ϯ 14. Tracing of the anten- appeared rotated by Ϸ45° along the anterior-posterior and nal nerve (AN) revealed that ORN axons separate (to- medio-lateral axes. In addition to the m- and l-ACT, and gether with axons from motor neurons and mechanorecep- similar to the situation described in the honeybee, three tors that proceed further to the dorsal lobe, DL) forming small ml-ACTs (1–3) projected directly to the lateral horn seven distinct sensory tracts that enter the AL (T1–T7). (LH), and two of them (ml-ACT 2 and 3) formed a “lateral Each tract innervated a characteristic cluster of glomeruli network” between the mushroom body (MB) vertical lobe dividing the AL into seven glomerular subregions. In the and the LH (Fig. 2F,G). following, these glomerular clusters are termed T1–7 clus- In the example shown, a total of 216 glomeruli in the ters (Figs. 1, 2B,C). An overview of the number of glomer- rostral-ventral hemisphere of the AL were innervated uli innervated by T1–7 together with their volumes and by PNs of the l-ACT (green; Fig. 2D,E; Table 1). The ACT associations are given in Table 1. The tracts were l-ACT axons ran dorsally through the inner nonglo- identified by closely following their trajectories image by merular core of the AL. They formed a single prominent image in 3D image stacks (see example of an image stack bundle leaving the AL in dorsal direction, bending in Suppl. material to Fig. 1, and a rotating image of the 3D rostral-laterally and projecting to the ipsilateral LH reconstruction in Suppl. materials to Fig. 2). T1 proceeded before innervating the MB calyces (Fig. 2E–G). The at the ventral surface of the AL and innervated a small somata of l-ACT PNs were arranged in two distinct cluster of Ϸ34 glomeruli in the ventral-rostral part of the clusters (lSC): a large ventral cluster on the l-ACT AL (Figs. 1A,B, 2B; Table 1). This cluster was flanked hemisphere (lSC1, Fig. 2E) and a smaller cluster along laterally by the T2 and medially by the T3 cluster. The T2 the rostral lateral rim of the l-ACT hemisphere (lSC 2; cluster extended along the ventral-lateral side of the AL visible in 3D Suppl. material). A total of 218 glomeruli and consisted of Ϸ56 glomeruli. The T3 cluster with 96 in the caudal-dorsal AL hemisphere were innervated by glomeruli spread along the ventral medial part of the AL. m-ACT PNs (magenta) (Table 1; Fig. 2D,E). The axons Since the clear separation of T3 and T5 glomeruli was of m-ACT PNs projected rostrally through the nonglo- most difficult, we used additional criteria to distinguish merular AL neuropil and formed two separate bundles. among the two clusters such as the smaller size of T3 These bundles merged after passing the l-ACT-bundle, glomeruli compared to T5 glomeruli. A group of Ϸ78 large but before exiting the AL to form the m-ACT (Fig. 2F,G). glomeruli comprised the T4 cluster at the dorsal-lateral Somata of the m-ACT (mSC) were arranged in two large 430

Fig. 1. Antennal sensory innervation of the antennal lobe (AL). glomerular clusters (for further details, see complete image stack in A–D: Series of confocal images of an AL with an anterogradely labeled Suppl. materials). The asterisk in D labels the single T7 glomerulus antennal nerve showing the sensory innervation of AL glomeruli via that is not innervated by m- or l-ACT PNs. Spatial directions indi- seven distinct antennal sensory tracts (T1–T7) and the associated cated in A: rostral, r; caudal, c; medial, m; lateral, l. Scale bar ϭ 100 cluster of glomeruli. Single optical sections at different depths (indi- ␮m in A (applies to all). cated in A–D) show the seven tracts and parts of the innervated clusters, one triangle-shaped cluster in the rostral me- staining belonged to the dorsal-rostral T7 cluster (as- dial region (mSC1) and a smaller band of somata in the terisk in Fig. 1D). It showed similarities regarding its caudal lateral part of the AL (mSC2, Fig. 2E). With the position with glomerulus D02 in the honeybee (Kirsch- exception of one glomerulus, all glomeruli within the AL ner et al., 2006). The relatively small variance in the could be assigned to either the m- or l-ACT, and none of total number of glomeruli (Ϯ14, see results above) indi- them was innervated by both (Table 1). The one glomer- cates that variance in the input-output connectivity in ulus that was never found labeled after m- or l-ACT the seven clusters is rather low. 431

Afferent and efferent connection of AL PNs leaving the AL via three ml-ACT showed a distinct glomeruli projection pattern in the lateral protocerebral lobe (LPL) and LH. The axons of the ml-ACT1 bypassed the LPL and To assign the antennal sensory tracts to either the m- or targeted the LH only. The axons and synaptic fields of l-ACT we combined anterograde ORN mass-fills with ei- ml-ACT 2 and 3 formed a characteristic arborization pat- ther retrograde labeling of the m-ACT (two preparations) tern within the LPL similar to the “lateral network” found or the l-ACT (three preparations). All preparations re- in the honeybee. Kirschner et al. (2006) describe three vealed a clear distribution of six of the seven tracts plus distinct olfactory innervation foci in the LPL of the hon- associated glomeruli cluster to either the m- or l-ACT eybee: the ring neuropil, located ventrally in the brain hemisphere. T1, T2, and T4 exclusively belonged to the around the vertical lobe of the MB, the triangle, a small l-ACT hemisphere (Table 1). T5 and T6 were innervated region in the dorsal center of the LPL, and the lateral by m-ACT PNs. Similarly, the T7 cluster was innervated bridge between the triangle and the LH. In C. floridanus by m-ACT PNs, except for one large glomerulus, which similar innervation foci of ml-ACT axons were found in was neither innervated by m- nor l-ACT PNs. The glomer- corresponding regions of the LPL (Fig. 3C,D). Compared uli of the T3 cluster could be divided into two about to the honeybee, the ring neuropil in C. floridanus ap- equally large subgroups innervated by either the m- or peared to extend less around the MB vertical lobe. l-ACT. The upper rostral part belonged to the l-ACT hemisphere, the caudal part to the m-ACT hemisphere. A sim- Calcium imaging of odor responses in the ilar innervation pattern was recently shown for the T3a antennal lobe cluster in the honeybee (Kirschner et al., 2006). We measured intracellular calcium changes in AL pro- Segregated olfactory input to the MB jection neurons (PNs) in response to different odors. Both m- and l-ACT PNs were retrogradely filled with the calyces and lateral protocerebral lobe calcium-sensitive dye Fura-2 dextran and successful dye Olfactory information from the AL is transferred to the uptake resulted in a bright fluorescence of glomeruli and MB calyces via the m- and l-ACT. Their course through somata clusters in the AL. In all preparations tested spon- the protocerebral lobe is shown in the projection view in taneous neuronal activity (changes in calcium amplitude) Figure 2F and a 3D visualization of their course is shown could be observed in areas that were comparable to single in Figure 2G. Double staining of the m- and l-ACT projec- glomeruli in shape and size (activity spots). During odor tions in the same preparation (n ϭ 1) revealed a promi- stimulation, changes in calcium responses were visible in nent innervation of the lip region (Fig. 2H). This pattern larger areas of the AL (activity areas) or in activity spots. was confirmed by single stainings of either the m- (n ϭ 3) In most cases it was not possible to assign activity areas to or l-ACT (n ϭ 4). The collar region was not innervated by particular glomeruli or activity spots to a single glomeru- m- or l-ACT PNs. Throughout the basal part and the lus because optical resolution of the fluorescent images did upper part of the lip only the peripheral part was inner- not allow reliable discrimination of all glomeruli (Ϸ50) in vated by PNs of the m-ACT, whereas the inner part was the field of view. Therefore, we compared odor represen- innervated by the l-ACT. The lip region appeared subdi- tation in the AL in response to stimulation using activity vided into three concentric layers (Fig. 2H; I, II, and III). areas. Glomerulus diameter in the imaged area ranged The inner core and the intermediate layer of the lip were between 15–20 ␮m (Fig. 2; volumes in Table 1). Some predominantly innervated by the l-ACT and only few bou- activity patterns covered larger areas within the AL, in- tons of m-ACT PNs were found. Compared to the inner dicating that in these cases groups of glomeruli rather core, the intermediate layer of the lip showed a higher than single glomeruli were activated. For measurements density of boutons. In contrast, the outer rim of the lip (see below), the ROIs were set not larger than 20 ␮mto region was most densely packed with boutons, exclusively avoid measurement of such “group” effects. from m-ACT PNs. Innervation of the LH was compartmentalized into pre- Odor-specific calcium signals dominantly m- or l-ACT domains (Fig. 3) (n ϭ 5 prepara- In response to stimulation, odor-specific activity areas tions). Figure 3A combines results from two different were found in 39 animals using 2–8 different odors at high preparations in two different ants. Figure 3A shows one concentration (dilution: 10Ϫ1). Repeated odor stimulation preparation with both m- and l-ACT stained and the with the same odor resulted in similar activity areas (data m-ACT innervation was analyzed and highlighted in the not shown). In response to some odors (citral, isoamylac- LH. Figure 3B shows a second preparation in which the etate, nerol, and heptanal) the activity areas recorded in m-ACT was cut to prevent the dye from diffusion to the individual ants covered large parts of the AL, whereas m-ACT target regions in the LH. In this case only l- and responses to other odors (1-octanol, 1-hexanol, ml-ACT fibers were stained, and their innervation pattern n-undecane, and nerolic acid) were more confined in a few in the LH shows clear separation from the m-ACT target activity spots (Fig. 4A–F). As an example, odor-specific region. The axons terminated in a club-shaped medial activity areas measured in one animal are shown as false- compartment of the LH, which was exclusively innervated color-coded images. Pheromone stimulation resulted in by m-ACT PNs (Fig. 3A; dashed line). The closely sur- qualitatively similar activity areas as compared to general rounding neuropil contained only a few m-ACT innerva- odors (Fig. 4F,K,O,P–U). Odor-specific activity areas were tion, but dense branches of l-ACT and ml-ACT PNs (n ϭ 4) similar across individuals as shown for stimulation with (Fig. 3B). The data indicate a zone of overlap of l- and heptanal (dilution: 10Ϫ1), isoamylacetate (IAA; dilution: ml-ACT fibers in an intermediate region within the LH. 10Ϫ1), citral (dilution: 10Ϫ1), and nerolic acid (dilution: This pattern indicates a topographical separation of m- 10Ϫ9) in two ants (Fig. 4G–O). The activity areas in re- and l-ACT input to the LH strikingly similar to that found sponse to different odors overlapped both within the group in the honeybee (Kirschner et al., 2006). of general odors as well as between general odors and 432

Figure 2 433

TABLE 1. Summary of the Afferent and Efferent Connection of Glomeruli T1 T2 T3 T4 T5 T6 T7 ⌺

A ⌺ glomeruli 34 56 96 78 36 128 6 434 B ⌺ volume 91.6 159.5 402.8 423.6 255.0 331.4 133.2 1,797.1 [␮m3 ϫ 103] C Max volume 10.0 8.4 9.8 19.2 12.6 7.0 41.2 [␮m3 ϫ 103] D Min volume 0.9 1.1 1.3 1.8 4.9 0.4 10.8 [␮m3 ϫ 103] E Mean volume 2.7 2.8 4.2 5.4 7.1 2.6 22.2 [␮m3 ϫ 103] F Output tract innervation l-ACT l-ACT l-ACT/m-ACT l-ACT m-ACT m-ACT m-ACT (5 out of 6)

A, B: Sum of the number and total volumes (in ␮m3 ϫ 103) of all glomeruli belonging to a cluster supplied by the seven antennal sensory tracts T1–T7. C, D, E: Maximum (Max), minimum (Min) and mean volume of glomeruli belonging to the different tract-specific clusters. F: Affiliation of the different clusters to the two main projection neuron output tracts (m- or l-ACT). pheromones (estimated areas ranged from very little over- concentration-dependent only at very high and very low lap to up to Ϸ70%). The activity areas of the two phero- concentrations (Fig. 4P–U). The response threshold for mones overlapped each other in the range of Ϸ10–70%. nerolic acid was at a dilution of 10Ϫ11 with only a single activity spot. At a dilution of 10Ϫ9, additional activity Odor concentration threshold for calcium spots were found (arrows in Fig. 4Q), and this pattern signals remained largely stable over a concentration range of 7–8 To investigate the sensitivity of the olfactory system in log units (Fig. 4Q–T). At the highest odor concentration C. floridanus we measured the lowest odor concentration tested (dilution: 10Ϫ1) a further activity spot added to the that elicited consistent calcium patterns in the AL (Fig. 5). existing pattern (Fig. 4U; arrow). We found a very high sensitivity of PN calcium responses The amplitude of the calcium signals from activity spots to both general odors (1-hexanol, 1-octanol) and phero- was analyzed over different odor concentrations (Fig. 5B). mones (nerolic acid, n-undecane). In at least 80% of the ROIs were defined around activity spots and response investigated animals, activity areas were found in re- amplitudes were normalized within each animal (see Ma- sponse to very low odor concentrations (dilution: 10Ϫ11)of terials and Methods). The amplitude of the calcium signal the four odors tested (Fig. 5A). We only tested one odor within the ROIs was concentration-dependent in some quality at different concentrations in each animal (total cases (ROIcdϩ) and not concentration-dependent in other ϭ n 29), thus, all four response thresholds are indepen- cases (ROIcdϪ). The pairs of ROIs, selected in each prep- dent of each other. aration of the seven animals investigated, were split to

either ROIcdϩ or ROIcdϪ regions. This was possible in ﬁve Dependence of odor representation on odor animals shown in Figure 5B; in the remaining two ani- concentration mals no clear concentration dependency was found. To

The trail pheromone component nerolic acid was used compare the duration of the calcium signal, ROIcdϩ and as stimulus to investigate the variance of activity areas ROIcdϪ regions were analyzed separately. As an example, and the dynamic range of the calcium activation patterns the time courses of the response of one cdϩ and one cdϪ to across odor concentrations (11 log units). In all seven three odor concentrations are shown together with the animals investigated we found that the spatial pattern of control (response to solvent) in Figure 5C. The response activity areas in response to nerolic acid were odor- dynamics for all measured ants were further analyzed.

Fig. 2. The olfactory pathway in the brain of Camponotus florida- rostral-ventral hemisphere of the AL, m-ACT glomeruli (magenta) nus. Afferent and efferent innervation of antennal lobe (AL) glomer- innervate the caudal-dorsal hemisphere of the AL. The somata of uli, course of the output tracts, and their projection in the mushroom- l-ACT neurons (dark green) form a large cluster at the ventral part of body (MB) calyx. A: 3D reconstruction of the major neuropils in the the AL (lSC1) and a smaller cluster at the rostral rim of the AL (lSC2; brain of a C. floridanus worker. Central complex, CC; optic lobe, OL; not visible in this view; see Suppl. online material). Somata of the mushroom body, MB. B,C: 3D reconstructions of the antennal-sensory m-ACT neurons (dark magenta) form two clusters, one in the rostral- tract specific innervation of glomerular clusters (T1–T7) based on 200 medial (mSC1) and one in the caudal-lateral region of the AL (mSC2). optical sections. B, ventral and C, dorsal view. The color code defines F: Projection view of an anterograde mass fill of the AL output tracts the different sensory tracts (T1–T7) and their innervated glomerular and their projections in the MB and LH. The m- and l-ACT run to the clusters. The T1 cluster (orange) contains Ϸ34 glomeruli, the T2 medial and lateral MB calyces (mCa, lCa) and to the LH. Three small cluster (green) Ϸ56 glomeruli, the T3 cluster (magenta) Ϸ96 glomer- ml-ACTs (1–3) project directly to the LH, two of them (ml-ACT 2 and uli, the T4 cluster (light blue) Ϸ78 glomeruli, the T5 cluster (light 3) form a “lateral network” (ln, yellow in G) between the MB vertical green) Ϸ36 glomeruli, the T6 cluster (dark blue) is the largest cluster with Ϸ128 glomeruli, and the T7 cluster (yellow) is the smallest lobe and the LH. G: 3D-reconstruction of the mass fill shown in F. The cluster with the six largest glomeruli in the AL (see Table 1). D: Ortho l-ACT is shown in green, the m-ACT is in magenta, and all three slice of a doubly stained AL showing the glomerular innervation of the ml-ACTs in orange. H: Double labeling of the m-ACT (magenta) and two output tracts, m- and l-ACT (magenta and green), within the AL. l-ACT (green) projections in the medial MB calyx showing a distinct Please note that the colors in D,E do not reflect connectivity with the olfactory innervation pattern within the mushroom body calyx lip and glomeruli with similar colors in B. E: Reconstruction of the AL shown collar region. The lip can be separated into three distinct layers in D. Ventral view showing the tract-specific innervation of glomeruli according to their innervation (I, II and III). See text for details. and the position of m- (magenta) and l-ACT (green) PN somata clus- Directions: rostral, r; caudal, c; ventral, v; dorsal, d. medial, m; lateral, ter. Glomeruli innervated via the l-ACT (green) are located in the l. Scale bars ϭ 100 ␮m in A–G; 50 ␮minH. 434

Fig. 3. Spatial organization of ACT projections within the proto- innervation areas of the l-ACT and ml-ACTs. C,D: Lateral network cerebrum. Projection views at different depths (indicated in A–D) within the LPL formed by PNs of the ml-ACT 2 and 3. Three different showing the innervation patterns of the ACTs within the lateral horn innervation foci can be distinguished: the ring neuropil (rn), triangle (LH) and the lateral network within the lateral protocerebral lobe (tr), and lateral bridge (lb). Lateral calyx, lC; medial calyx, mC; (LPL). A: Club-shaped target region (dashed line) of the m-ACT vertical lobe, vL. Directions: rostral, r; caudal, c; ventral, v; dorsal, d. within the LH. B: Target region of the l-ACT and the ml-ACTs within medial, m; lateral, l. Scale bars ϭ 100 ␮m. the LH. The m-ACT region (dashed line) is clearly separated from the

Increasing odor concentrations evoked a higher amplitude DISCUSSION of the calcium signal in ROIcdϩ regions and also a longer duration of the calcium signal (Spearman’s rank correla- This study represents a first comprehensive structure– tion, R ϭ 0.69, P Ͻ 0.05, slope ϭ 0.36). Although increas- function analysis within the central olfactory system of an ing odor concentration did not result in a higher ampli- ant’s brain. Our structural data revealed that the AL of C. tude of the calcium signals in ROIcdϪ regions, the duration floridanus contains a comparatively high number of olfac- of the calcium signal was also concentration-dependent tory glomeruli (up to 464) supplied via seven distinct an- (Spearman’s rank correlation, R ϭ 0.50, P Ͻ 0.05, slope ϭ tennal sensory tracts. At the output side the AL is subdi- 0.33). vided into two almost equally sized hemispheres 435 regarding glomeruli innervated by one of two prominent, and Bru¨ckner, 2001; Sandoz 2006). This, together with the presumably uniglomerular PN output tracts, the m- and fact that in our calcium imaging experiments the response l-ACT. The projections of m- and l-ACT PNs run in a patterns to stimulation with pheromonal and nonphero- reverse order to the MB calyces and the LH, innervating monal substances were equally sensitive and not obvi- different compartments. In addition, three small tracts ously segregated indicates that processing of pheromonal (ml-ACT1–3) project through the LPL toward the LH, two and nonpheromonal odors may be combinatorial in the ant of them with side branches in the LPL (ml-ACT2,3) form- AL (see further discussion of this aspect below). ing a lateral network with several distinct olfactory foci Does the high number of glomeruli in the AL of C. (ring neuropil, triangle, and lateral bridge). floridanus indicate more sophisticated olfactory process- Functional calcium imaging studies of PN activity re- ing compared to, e.g., the honeybee, the fly, or moth? vealed reproducible glomerular activation patterns in re- Comparison of the number of odorant receptor (OR) genes sponse to pheromonal (alarm, trail) and general odors at in recently sequenced insect genomes of Drosophila mela- about equally high sensitivity. No obvious spatial segre- nogaster (Ϸ62; Clyne et al., 1999; Robertson et al., 2002; gation among pheromonal and nonpheromonal odor acti- Vosshall et al., 2002), the malaria mosquito Anopheles vation patterns was observed for the odors tested. gambiae (Ϸ79; Hill et al., 2002), and the honeybee (Ϸ170; Whereas the spatial response patterns to stimulation with Weinstock et al., 2006) reveal a rough correlation between trail pheromone (nerolic acid) were remarkably stable the number of OR genes and the number of olfactory over a wide range of intensities (Ϸ8 log units), the re- glomeruli in the AL (43 in Drosphila; Laissue et al., 1999; sponse intensities and especially response durations were 61 in Anopheles; Ghaninia et al., 2007; 164 in Apis; Galizia dependent of odor intensities. et al., 1999b; Kirschner et al., 2006). Similar relationships were also found in vertebrate olfactory systems (Buck and Antennal lobe design Axel, 1991). If we assume a similar correlation of OR The total number of AL glomeruli in C. floridanus is numbers and AL glomeruli in C. floridanus, we expect a almost 3-fold higher than in the honeybee (464 vs. 164) substantially higher number of ORs compared to the hon- and glomeruli in the AL are innervated and organized into eybee. To further prove this hypothesis would require seven clusters innervated by seven antennal sensory mapping OR expression to find out whether it may lead to tracts compared to four main tracts (T1–4) in the honey- novel glomeruli. The process of glomerulus induction it- bee (Abel et al., 2001), or six input tracts when subdivi- self, however, is not likely to be directly influenced by ORs sions of T3 (a–c) are counted as separate tracts as sug- because of their late expression in insects (Vosshall et al., gested by Kirschner et al. (2006). Despite substantial 2002). C. floridanus has small optic ganglia compared to differences in the antennal sensory input the general in- relatively large ALs (Fig. 2A). In contrast, the honeybee nervation pattern of two prominent output tracts (m- and has elaborated optic ganglia and sophisticated visual ca- l-ACT) is very similar compared to the situation in the pabilities (Gronenberg, 2001; Stach et al., 2004).In this honeybee (Kirschner et al., 2006). The dorsal-rostral part context, evolution of a higher number of glomeruli in of the AL in C. floridanus is innervated only by l-ACT PNs, highly olfactory ants appears plausible. In the same line, the ventral-caudal part exclusively by m-ACT PNs. Com- neuroanatomical studies of the more primitive ponerine pared to the honeybee, both hemispheres are rotated by ant, Harpegnathos saltator, a highly visual predator, re- about 45°C. The division of the AL into two hemispheres vealed a much lower number of Ϸ178 glomeruli (Hoyer et with each half containing about 50% of the glomeruli very al., 2005). An interesting hypothesis to follow up in the likely reflects a typical feature of the olfactory pathway in future is that the evolution of new OR genes may come Apocrita (Kirschner et al., 2006). Whether it is a common along with the formation of new sensory tracts and clus- feature of all hymenoptera needs to be shown in future ters of glomeruli in the AL and/or the increase/decrease in neuroanatomical studies of other advanced and basal hy- the size of existing ones. Future comparison of more menoptera. closely related ant species of the subfamily Formicinae Since our calcium imaging data gives only information and future genomic data may reveal further insight into about the physiological properties of glomeruli belonging the question. to clusters T1–3 on top of the AL, other methods that allow Structural comparison of sensory-tract-specific glomer- imaging of deeper layers of the AL, such as two photon ular clusters in C. floridanus indicates some similarities confocal imaging or electrophysiology studies, are re- (possibly homologies), but also clear differences compared quired to gain information about the functional role of the to the situation in the honeybee (Kirschner et al., 2006). division of the AL into m- and l-ACT hemispheres. It will The T7 cluster of C. floridanus shares striking similarities be especially interesting to see whether a common picture with the T4 cluster in the honeybee: both possess six large may emerge from a physiological comparison between the glomeruli at the dorsal end of the AL with a unique sen- honeybee and the ant. In addition to mapping different sory innervation pattern and with one or two glomeruli categories of odors, the function of a dual pathway could as neither connected to the m-ACT nor to the l-ACT. If the well be to code different temporal aspects of the stimulus, two clusters turn out to be homologous, the perseverance as suggested by Mu¨ller et al. (2002), or to extract different of their anatomy throughout the Apocrita lineage would stimulus parameters. To study temporal aspects of odor predict an important functional significance of these six coding, comparative electrophysiological studies of re- glomeruli. Physiological, neurochemical, and molecular sponse profiles are needed. characterization of this deep cluster, therefore, will be We did not find any obviously enlarged glomeruli com- very elucidating. parable to the macroglomerulus found at the AL entrance The T3 cluster of C. floridanus shares similarities with in leaf-cutting ants (Kleineidam et al., 2005) or macroglo- the T3a cluster after Kirschner et al. (2006), especially its merular complexes in moths or in honeybee drones (e.g., position at the ventral-medial AL and the division of its Hildebrand et al., 1997; Ro¨ssler et al., 1999a; Brockmann glomeruli to both ACT-hemispheres. The best candidates 436

Figure 4 437 for being homologous to the honeybee T1 cluster are T1, 2, different odors. We further were able to show that odor- and4inC. floridanus: all three together make up the specific activity patterns are conserved between individu- rostral-lateral part of the AL, like in the honeybee if one als, similar to findings in the honeybee (Galizia et al., considers the 45° shift of the ant AL, and they all belong to 1999b). the l-ACT hemisphere. On the opposite side of the AL, the Only in some cases could calcium signals from the PNs large T6 cluster in C. floridanus shares five anatomical be assigned to individual glomeruli (activity spots). Large features with the very small T3b cluster in the honeybee activity regions indicate that the odor used for stimulation (after Kirschner et al., 2006): first a position at the medial- was represented in several adjacent glomeruli rather than caudal AL flanking the AN, second a sensory tract that in isolated single glomeruli of the AL. Studies in the approaches the glomeruli from the periphery of the AL, honeybee indicate that adjacent glomeruli may have sim- third a clear separation from the other glomeruli clusters, ilar response profiles (Sachse et al., 1999). In our study the fourth a composition of relatively small glomeruli, and size of cohesive activity regions covering several glomeruli fifth the affiliation to the m-ACT hemisphere. The most remained constant over a large range of concentrations, striking difference of the T6 cluster in C. floridanus com- and in most cases was only marginally smaller at odor pared to T3b cluster in the honeybee is the high number of concentration threshold. Are the large activity regions the Ϸ128 glomeruli indicating that this could be a spot of result of scattered fluorescence from activated glomeruli strong evolutionary proliferation of glomeruli in the out of focus, e.g., in deeper layers of the AL? Glomeruli in ant AL. the AL of C. floridanus are clustered (Figs. 1, 2), and some glomeruli may be above or below the focal plane during Odor representation in the AL calcium imaging. Repeated stimulation with the same Calcium imaging in the AL revealed odor-induced activ- odor at different focal planes revealed large activity re- ity of PNs. We employed the retrograde loading technique, gions also in z-direction, and no confined activity spots originally developed in the honeybee by Sachse and Gali- were found up to a depth of 80 ␮m (data not shown). This zia (2002), for the first time successfully in the ant. Al- indicates a large or even complete overlap of response though both l- and m-ACT PNs were backfilled with our profiles in at least some neighboring glomeruli. Other technique, our neuroanatomical analyses show that the glomeruli seemed to have a more distinct response profile field of view in calcium imaging experiments was re- compared to their adjacent glomeruli. For example, with stricted to the ventral part of the AL almost exclusively an increasing concentration of nerolic acid, some activity innervated by l-ACT PNs (with only a few exceptions of spots appeared at low concentration and remained con- m-ACT glomeruli in the T3 cluster; see Fig. 2E). There- stant in size up to high odor concentrations. In this case, fore, all further conclusions are mostly limited to l-ACT no glomeruli next to the activity spots were recruited to PNs. the activation pattern. In our study we used only single- Odor-specific activation patterns in primary olfactory component odors, and it remains to be shown whether centers were described for a number of invertebrate and odor mixtures may elicit a more distinct activation pattern vertebrate species (e.g., Rodrigues, 1988; Friedrich and across glomeruli within the same activity region by pro- Korsching, 1997; Joerges et al., 1997; Rubin and Katz, cessing of odor information in the AL network. 1999; Hansson et al., 2003; Carlsson et al., 2005), but only The activity patterns in response to (nonsex) phero- one earlier study in a closely related Camponotus species mones (or pheromone components) were distributed (Galizia et al., 1999a). Activation patterns in response to across the field of view, and there was no obvious quali- different odors were overlapping, indicating that some tative difference compared to the activation patterns in glomeruli contribute to the activation pattern elicited by response to general odors. Even very low pheromone concentrations elicited a distributed activation pattern in the AL and no distinct cluster of pheromone specific glomeruli was found. Yamagata et al. (2006) using electrophysiological recording and staining of AL neurons described Fig. 4. Projection-neuron calcium responses in the antennal lobe. A: Fura-2 raw fluorescence image of the antennal lobe (AL). The “alarm pheromone-sensitive” glomeruli in another Cam- position of the antennal nerve (AN) and the spatial directions are ponotus species that form a cluster in the dorsal most part indicated. The spatial directions are valid for all false-color coded of the AL. This particular glomerular cluster was not images. B–F: False-color coded images showing activity areas in the accessible in our imaging experiments. Nevertheless, in AL of C. floridanus in response to different odors. The calcium signals contrast to this our calcium imaging experiments revealed were recorded from PNs of the AL loaded with Fura-2 dextran. 500 nl Ϫ a rather distributed representation of n-undecane in C. (dilution 10 1) for all odors. The different odors elicit odor-specific activity areas within one ant. G–O: The same odors elicit similar floridanus, and we conclude that for n-undecane as well as activity areas in different ants. Examples showing activity areas in for nerolic acid pheromone-specific clusters of glomeruli do response to heptanal (dilution of 10Ϫ1), isoamylacetate (IAA; dilution not exist in the AL of workers. This confirms and extends Ϫ Ϫ Ϫ of 10 1), citral (dilution of 10 1), nerolic acid (dilution of 10 9)intwo similar observations from a calcium imaging study in different ants. P–U: Response patterns to the trail pheromone com- Camponotus rufipes (Galizia et al., 1999a), and is consis- ponent nerolic acid presented at different concentrations. High concentration leads to the recruitment of additional activation spots tent with the rather distributed representation of nonsex (arrows in Q and U). The dilutions are given in the images. Note that pheromones in glomeruli within the AL of the honeybee the activity areas in response to general odors covered large parts of (Joerges et al., 1997; Galizia et al., 1999c). It remains to be the AL (B,H,M, citral; E, nerol; G,L, heptanal) or were more confined shown whether this arrangement is advantageous or even with only few activity spots (C, 1-hexanol; D, 1-octanol). Pheromones necessary to allow context-specific behavioral responses to (F, the alarm pheromone n-undecane; K,O,P–U, the trail pheromone pheromones as described for ants (Ho¨lldobler and Wilson, component nerolic acid) elicited qualitative similar activity areas compared to general odors. False-colors are scaled in images B–F as 1990; Knaden and Wehner, 2003). Differential neuro- shown in B, images G–O as shown in G, and images in P–U as shown modulatory innervation of the AL (Dacks et al., 2006; in P. caudal, c; lateral, l; medial, m; rostral, r. Scale bars ϭ 100 ␮m. Ziegler et al., 2007) are likely to promote primary process- 438

Fig. 5. Sensitivity of odor responses and intensity dependence of sponses to nerolic acid in five concentration-dependent (cdϩ) and five responses to nerolic acid. A: Response-threshold odor concentration concentration-independent (cdϪ) ROIs within the antennal lobe of for pheromones (NA, nerolic acid; UN, n-undecane) and general odors five animals across a range of odor concentrations (dilution: 10Ϫ12 to (HE, 1-hexanol; OC, 1-octanol). Percentage of animals in which a 10Ϫ1). C: Temporal dynamics of the calcium responses to nerolic acid calcium signal (activity areas) could be measured in response to in one concentration-dependent (cdϩ) and one concentration- different odor concentrations (dilutions: white bars, 10Ϫ12; gray bars, independent (cdϪ) ROI within the antennal lobe of one animal at 10Ϫ11; black bars 10Ϫ10). Loading the odor cartridge with 10Ϫ11 dilu- three different odor concentrations (10Ϫ11,10Ϫ5, and 10Ϫ2) and the tion of any of the four odors was sufficient to elicit a calcium response control stimulus (solvent only). Threshold level of 40% is indicated by in at least 80% of the animals (n, number of animals tested). B: Nor- the horizontal dotted line. The response duration was longer at higher malized amplitudes (see Materials and Methods) of the calcium re- odor concentrations compared to lower concentrations. 439 ing of pheromone information and certainly need further tudes at the highest concentration than at any other con- investigation. centration (Fig. 5B). In case of high basal intracellular The concentration thresholds we measured were at ex- calcium concentration, weak neural responses may al- tremely low concentrations. Odor loading of the filter pa- ready lead to a saturation of the Fura-2 signal. In these per in the range of 0.05 pg (equivalent to dilutions of cases the calcium amplitude would not increase with in- 10Ϫ11) was sufficient to elicit reliable calcium amplitudes creasing response strength, but the duration would. How- in PNs, and concentration thresholds were similar for ever, the fact that even in cdϪ cases we observed a slight both pheromones and general odors. Although it is diffi- increase in the amplitude only at the highest concentra- cult to compare data from different laboratories using tion indicates that the Fura-2 signal was not saturated. In different types of odor stimulation devices (e.g., variation contrast to variations with respect to graded amplitudes, in air flow velocity and volume), the odor sensitivity in we found graded response durations in all cases. This ants seems to be 5 log units higher compared to what was indicates that response duration might be an important described for the honeybee (Sachse and Galizia, 2003). In parameter for assessing odor concentration, although one addition, in the honeybee the spatial activation pattern has to keep in mind that the duration of the Fura-2 signal was shown to be concentration invariant only over a lim- may be strongly influenced by both the strength (spike ited range of concentrations. Our results in C. floridanus rate) and the duration of a neural response. Further elec- indicate that the spatial activation pattern in response to trophysiological analyses at the single neuron level are the trail pheromone component is largely invariant over a required to fully clarify this aspect in odor coding. rather large concentration range of Ϸ8 log units. Only at Odor information is relayed to higher brain regions via the lowest and highest concentrations, glomeruli disap- a combinatorial code of PN activation. Although it is not peared or additional ones were recruited. Concentration- fully understood how odor quality and intensity are coded dependent activity patterns were also described in other by the population of PNs, it seems plausible that the species (Rubin and Katz, 1999; Johnson and Leon, 2000; temporal structure of PN activity is important. Mixtures Meister and Bonhoeffer, 2001, Fu␤ and Korsching, 2001; of odors lead to a synchronization of pheromone-sensitive Wachowiak and Cohen, 2001). The main reason for the PNs in moths (Lei et al., 2002), and in the locust oscilla- change in activity patterns at the level of the AL or olfactory synchronization of activity is considered as an impor- tory bulb probably is that the receptive range of individual tant feature for odor coding and recognition (Laurent, ORNs is concentration-dependent and broadens with in- 2002). Prolonged activation of PNs in response to higher creasing odor concentrations (Firestein et al., 1993; odor concentrations might enhance temporally coordi- Duchamp-Viret et al., 2000; de Bruyne et al., 2001). The nated PN activity, and temporal coordination may be im- perceived odor quality can depend on odor concentration, portant for both odor quality and intensity coding (Lei et a phenomenon known in human psychophysics (Gross- al., 2004). In this respect, spatial separation of m- and Isserhoff et al., 1988) and from experiments in Drosophila l-ACT PN pathways may be an important addition in the (Rodrigues and Siddiqi, 1983; Stensmyr et al., 2003). In Apocritan lineage reflecting the evolution of advanced pro- ants, concentration-dependent responses may result, e.g., cessing of a large variety of both chemical communication in a well-defined behavioral sequence when approaching a signals and environmental odors. (alarm) pheromone source (active space hypothesis; Brad- shaw and Howse, 1984; Ho¨lldobler and Wilson, 1990). The idea that indeed the activity pattern found in the AL ACKNOWLEDGMENTS corresponds to perceptual measures recently received We thank Wulfila Gronenberg for help with neuroana- strong support from a study in the honeybee (Guerrieri et tomical tracing techniques, Malu Obermayer for expert al., 2005). Interestingly, in response to trail pheromone we technical assistance, Mona Alzheimer for 3D- found activity spots only adding to the pattern at very reconstruction assistance, and Annett Endler for help high or low concentrations. Sampling of a chemical trail with rearing of Camponotus floridanus. with antennation movements may cause large fluctua- tions at ORNs; therefore, it may be important to maintain a constant spatial pattern at the level of the PN output in LITERATURE CITED order to stabilize odor quality perception. Higher concen- Abel R, Rybak J, Menzel R. 2001. Structure and response patterns of trations of the trail pheromone, on the other hand, en- olfactory interneurons in the honeybee, Apis mellifera. J Comp Neurol hance the probability of a worker to follow a trail. There- 437:363–383. fore, both the maintenance of a constant quality Bradshaw JWS, Howse PE. 1984. Sociochemicals of ants. In: Bell WJ, perception and, at the same time, assessing odor intensity Carde RT, editors. Chemical ecology of insects. London: Chapman & is important for trail-following behavior. Our results indi- Hall. p 429–473. cate that the response duration may be a reliable measure Brockmann A, Bru¨ckner D. 2001. Structural differences in the drone olfactory system of two phylogenetically distant Apis species, A. florea for coding odor intensity. Concentration-invariant activa- and A. mellifera. Naturwissenschaften 88:78–81. tion patterns together with graded calcium-signal ampli- Buck L, Axel R. 1991. A novel multigene family may encode odorant tudes, as also described in the honeybee, led to the as- receptors: a molecular basis for odor recognition. Cell 65:175–187. sumption that the activity pattern is the important Carlsson MA, Hansson BS. 2003. Dose-response characteristics of glomer- parameter for assessing odor quality while the peak activ- ular activity in the moth antennal lobe. Chem Senses 28:269–278. ity of the PNs relates to odor intensity (Galizia and Men- Carlsson MA, Galizia CG, Hansson BS. 2002. Spatial representation of zel, 2001). For some activity regions, but not for all, we odours in the antennal lobe of the moth Spodoptera littoralis (Lepidop- tera: Noctuidae). Chem Senses 27:231–244. found graded calcium-signal amplitudes with increasing Carlsson MA, Knusel P, Verschure PFMJ, Hansson BS. 2005. Spatio- odor concentrations. In other cases, activity regions with temporal Ca 2ϩ dynamics of moth olfactory projection neurones. Eur almost constant calcium-signal amplitudes over a wide J Neurosci 22:647–657. range of concentrations had considerably stronger ampli- Christensen T, Hildebrand J. 1987. Male-specific, sex-pheromone-selective 440

projection neurons in the antennal lobes of the moth Manduca sexta. Johnson BA, Leon M. 2000. Modular representations of odorants in the J Comp Physiol A 160:553–569. glomerular layer of the rat olfactory bulb and the effects of stimulus Clyne P, Certel S, de Bruyne M, Zaslavsky L, Johnson W, Carlson J. 1999. concentration. J Comp Neurol 422:496–509. The odor specificities of a subset of olfactory receptor neurons are Kirschner S, Kleineidam CJ, Zube C, Rybak J, Gru¨newald B, Ro¨ssler W. governed by Acj6, a POU-domain transcription factor. Neuron 22:339– 2006. Dual olfactory pathway in the honeybee, Apis mellifera. J Comp 347. Neurol 499:933–952. Dacks AM, Christensen TA, Hildebrand JG. 2006. Phylogeny of a Kleineidam CJ, Obermayer M, Halbich W, Ro¨ssler W. 2005. A macroglo- serotonin-immunoreactive neuron in the primary olfactory center of merulus in the antennal lobe of leaf-cutting ant workers and its possi- the insect brain. J Comp Neurol 498:727–746. ble functional significance. Chem Senses 30:383–392. de Bruyne M, Foster K, Carlson JR. 2001. Odor coding in the Drosophila Knaden M, Wehner R. 2003. Nest defense and conspecific enemy recogni- antenna. Neuron 30:537–552. tion in the desert ant Cataglyphis fortis. J Insect Behav 6:717–730. Duchamp-Viret P, Duchamp A, Chaput MA. 2000. Peripheral coding in the Lahav S, Soroker V, Hefetz A, Vander Meer RK. 1999. Direct behavioral rat and frog: quality and intensity specification. J Neurosci 20:2383– evidence for hydrocarbons as ant recognition discriminators. Naturwis- 2390. senschaften 86:246–249. Firestein S. 1993. Olfactory receptor neuron signaling. ACS Sym Ser Laissue PP, Reiter C, Hiesinger PR, Halter S, Fischbach KS, Stocker RF. 825:110–122. 1999. Three-dimensional reconstruction of the antennal lobe in Dro- Friedrich RW, Korsching SI. 1997. Combinatorial and chemotopic odorant sophila melanogaster. J Comp Neurol 405:543–552. coding in the zebrafish olfactory bulb visualized by optical imaging. Laurent G. 2002. Olfactory network dynamics and the coding of multdi- Neuron 18:737–752. mensional signals. Nat Rev Neurosci 3:884–895. ␤ Fu SH, Korsching SI. 2001. Odorant feature detection: Activity mapping Lei H, Christensen TA, Hildebrand JG. 2002. Local inhibition modulates of structure response relationship in the zebrafish and olfactory bulb. odor-evoked synchronization of glomerulus-specific output neurons. J Neurosci 21:8396–8401. Nat Neurosci 5:557–565. Galizia CG, Menzel R. 2001. The role of glomeruli in the neural represen- Lei H, Christensen TA, Hildebrand JG. 2004. Spatial and temporal orga- tation of odors: results from optical recording studies. J Insect Physiol nization of ensemble representations for different odor classes in the 47:115–130. moth antennal lobe. J Neurosci 24:11108–11119. Galizia CG, Menzel R, Ho¨lldobler B. 1999a. Optical imaging of odor-evoked Lenoir A, Fresneau D, Errard C, Hefetz A. 1999. Individuality and colonial glomerular activity patterns in the antennal lobes of the ant Campono- identity in ants: the emergence of the social representation concept. In: tus rufipes. Naturwissenschaften 86:533–537. Detrain C, Deneubourg JL, Pasteels JM, editors. Information process- Galizia CG, Sachse S, Rappert A, Menzel R. 1999b. The glomerular code for ing in social insects. Berlin: Birkhaüser. p 219–237. odor representation is species specific in the honeybee Apis mellifera. Nat Neurosci 2:473–478. Malun D, Waldow U, Kraus D, Boeckh J. 1993. Connections between the deutocerebrum and the protocerebrum, and neuroanatomy of several Galizia GC, McIlwrath SL, Menzel R. 1999c. A digital three-dimensional classes of deutocerebral projection neurons in the brain of male atlas of the honeybee antennal lobe based on optical sections acquired Periplaneta americana. J Comp Neurol 329:143–162. by confocal microscopy. Cell Tissue Res 295:383–394. Meister M, Bonhoeffer T. 2001. Tuning and topography in an odor map on Ghaninia M, Hansson BS; Ignell R. 2007. The antennal lobe of the African the rat olfactory bulb. J Neurosci 21:1352–1360. malarian mosquito, Anopheles gambiae-innervation and three- dimensional reconstruction. Arthropod Struct Dev 36:29–39. Mombaerts P, Wang F, Dulac C, Chao SK, Nemes A, Mendelsohn M, Edmonson J, Axel R. 1996. Visualizing an olfactory sensory map. Cell Gronenberg W. 1999. Modality specific segregation of input to ant mush- 87:675–686. room bodies. Brain Behav Evol 54:85–95. Mu¨ller D, Abel R, Brandt R, Zockler M, Menzel R. 2002. Differential Gronenberg W. 2001. Subdivisions of hymenopteran mushroom body caly- parallel processing of olfactory information in the honeybee, Apis mel- ces by their afferent supply. J Comp Neurol 435:474–489. lifera L. J Comp Physiol A 188:359–370. Gross-Isseroff R, Lancet D. 1988. Concentration-dependent changes of perceived odor quality. Chem Senses 13:191–204. Peele P, Ditzen M, Menzel R, Galizia CG. 2006. Appetitive odor learning does not change olfactory coding in a subpopulation of honeybee an- Guerrieri F, Schubert M, Sandoz JC, Giurfa M. 2005. Perceptual and tennal lobe neurons. J Comp Physiol A 192:1083–1103. neural olfactory similarity in honeybees. PLOS Biol 3:718–732. Robertson HM, Warr CG, Carlson JR. 2003. Molecular evolution of the Haak U, Ho¨lldobler B, Bestmann HJ, Kern F. 1996. Species-specificity in insect chemoreceptor gene superfamily in Drosophila melanogaster. trail pheromones and dufour’s gland contents of Camponotus atriceps Proc Natl Acad SciUSA100:14537–14542. and C. floridanus (Hymenoptera: Formicidae). Chemoecology 7:85–93. Rodrigues V. 1988. Spatial coding of olfactory information in the antennal Hansson BS, Anton S. 2000. Function and morphology of the antennal lobe: lobe of Drosophila melanogaster. Brain Res 453:299–307. new developments. Annu Rev Entomol 45:203–231. Rodrigues V, Siddiqi O. 1978. Genetic analysis of chemosensory pathway. Hansson BS, Christensen TA, Hildebrand JG. 1991. Functionally distinct Proc Ind Acad Sci 87B:147–160. subdivisions of the macroglomerular complex in the antennal lobe of the male sphinx moth Manduca sexta. J Comp Neurol 312:264–278. Rospars JP, Hildebrand JG. 2000. Sexually dimorphic and isomorphic Hansson BS, Ljundberg H, Hallberg E, Lofstedt C. 1992. Functional spe- glomeruli in the antennal lobes of the sphinx moth Manduca sexta. cialization ot olfactory glomeruli in a moth. Science 256:1313–1315. Chem Senses 25:119–129. Hansson BS, Carlsson MS, Kalinova B. 2003. Olfactory activation patterns Ro¨ssler W, Randolph PW, Tolbert LP, Hildebrand JG. 1999a. Axons of in the antennal lobe of the sphinx moth, Manduca sexta. J Comp olfactory receptor cells of transsexually grafted antennae induce devel- Physiol A 189:301–308. opment of sexually dimorphic glomeruli in Manduca sexta. J Neurobiol 38:521–541. Hildebrand JG, Shepherd GM. 1997. Mechanisms of olfactory discrimination: converging evidence for common principles across phyla. Annu Ro¨ssler W, Oland LA, Higgins MR, Hildebrand JG, Tolbert LP. 1999b. Rev Neurosci 20:595–631. Development of a glia-rich axon-sorting zone in the olfactory pathway of the moth Manduca sexta. J Neurosci 19:9865–9877. Hill CA, Fox AN, Pitts RJ, Kent LB, Tan PL, Chrystal MA, Cravchik A, Collins FH, Robertson HM, Zwiebel LJ. 2002. G-protein-coupled recep- Rubin BD, Katz LC. 1999. Optical imaging of odorant representations in tors in Anopheles gambiae. Science 298:176–178. the mammalian olfactory bulb. Neuron 23:499–511. Ho¨lldobler B. 1995. The chemistry of social regulation: multicomponent Sachse S, Galizia CG. 2003. The coding of odour-intensity in the honeybee signals in ant societies. Proc Natl Acad SciUSA92:19–22. antennal lobe: local computation optimizes odour representation. Eur Ho¨lldobler B, Wilson EO. 1990. The ants. Cambridge, MA: Harvard Uni- J Neurosci 18:2119–2132. versity Press. Sachse S, Rappert A, Galizia CG. 1999. The spatial representation of Hoyer SC, Liebig J, Ro¨ssler W. 2005. Biogenic amines in the ponerine ant chemical structures in the AL of honeybees: steps towards the olfactory Harpegnathos saltator: serotonin and dopamine immunoreactivity in code. Eur J Neurosci 11:3970–3982. the brain. Arthropod Struct Dev 34:429–440. Sandoz JC. 2006. Odour-evoked responses to queen pheromone compo- Joerges J, Ku¨ttner A, Galizia GC, Menzel R. 1997. Representation of nents and to plant odours using optical imaging in the antennal lobe of odours and odour mixtures visualized in the honeybee brain. Nature the honey bee drone Apis mellifera L. J Exp Biol 209:3587–3598. 387:285–288. Sandoz JC, Galizia CG, Menzel R. 2003. Side-specific olfactory conditioning 441

leads to more specific odor representation between sides but not within Vosshall LB. 2002. How the brain sees smells. Dev Cell 1:588–590. sides in the honeybee antennal lobes. Neuroscience 120:1137–1148. Vosshall LB, Wong AM, Axel R. 2000. An olfactory sensory map in the fly Singer TL. 1998. Roles of hydrocarbons in the recognition systems of brain. Cell 102:147–159. insects. Am Zool 38:394–405. Wachowiak M, Cohen LB. 2001. Representation of odorants by receptor Stach S, Bernard J, Giurfa M. 2004. Local-feature assembling in visual neuron input to the mouse olfactory bulb. Neuron 32:723–735. pattern recognition and generalization in honeybees. Nature 429:758– Wang JW, Wong AM, Flores J, Vosshall LB, Axel R. 2003. Two-photon 761. calcium imaging reveals an odor-evoked map of activity in the fly brain. Stensmyr MC, Giordano E, Balloi A, Angioy AM, Hansson BS. 2003. Novel Cell 112:271–282. natural ligands fro Drosophila olfactory receptor neurons. J Exp Biol 206:715–724. Weinstock GM, Robinson GE, Gibbs RA et al. 2006. Insights into social insects from the genome of the honeybee Apis mellifera. Nature 443: Strausfeld NJ. 2002. Organization of the honey bee mushroom body: Rep- 931–949. resentation of the calyx within the vertical and gamma lobes. J Comp Neurol 450:4–33. Xu FQ, Greer CA, Shepherd GM. 2000. Odor maps in the olfactory bulb. J Comp Neurol 422:489–495. Vassar R, Chao SK, Sitcheran R, Nunez JM, Vosshall LB, Axel R. 1994. Topographic organization of sensory projections to the olfactory bulb. Yamagata N, Nishino H, Mizunami M. 2006. Pheromone-sensitive glomer- Cell 79:981–991. uli in the primary olfactory centre of ants. Proc Biol Sci 273:2219–2225. Vickers NJ, Christensen TA, Hildebrand JG. 1998. Combinatorial odor Ziegler C, Starke NK, Zube C, Kirschner S, Ro¨ssler W. 2007. Neuromodu- discrimination in the brain: attractive and antagonist odor blends are lation and synaptic plasticity within olfactory centers in the brain of represented in distinct combinations of uniquely identifiable glomeruli. the carpenter ant, Camponotus floridanus. Proc 31st Go¨ttingen Neu- J Comp Neurol 400:35–56. robiology Conference; p 131.